AI models like GPT-5.2 Pro and Gemini 3.0 have less than 2% success on hard math problems, showing they create fake answers when they don't know.#AIHallucination, #GPT5, #Gemini3, #MathAI, #AIResearchhttps://newsletter.tf/ai-math-hallucination-unsolvable-problems-gpt-5-2/
